Did you mean that the amplitude of the sawtooth can change by up to 1 volt (due to imperfections in the differentiator and amplitude compensator)? And if the PWM is based on comparing the sawtooth with a fixed voltage, and the sawtooth amplitude changes with pitch, does the PWM width change with pitch?

You're correct. The pulse-width of the saw wave is very dependent on the amplitude of the saw, so when there's a 1 volt deficit in the saw wave's amplitude you lose about 8% of the pulse width. It's really not that noticeable in practice, though, and if anything it adds to the sonic character of the juno.